Testing for heteroscedasticity in instrumental variables approach with cross-sectional data set

[Jordan Smith]

I have a cross-sectional data set and I am using an instrumental variables approach.

I do not know whether to use ivregress 2sls or ivregress gmm.

The confusion comes from how to test which one to use. I have ran OLS and then completed a hettest. This has shown me that I have heteroscedasticity... but I am aware that this could be because of omitted relevant variables (as I argue in my paper that these exist for my specification hence IV). Now I have instrumented, is there a test I can do to check for heteroscedasticity. And am I right in believing in an IV regression the fix to this is to run ivregress gmm.

N.B. I have 1 endogenous regressor and 1 instrument.

Thank you in advance.

[Phil Bromiley]

You can easily use robust standard errors that take care of heteroscedasticity - ivregress and ivreg2 both allow for them.